Re: [Anuga-user] ANUGA

[Gareth D. <gar...@gm...>]

Hi Rupal,

Boundary conditions are a somewhat complex topic, where both 
mathematical and numerical issues interact.

It's not actually correct to say that if you want outflow, then you can 
use either transmissive or dirichlet. Mathematically there are 
additional requirements. An example paper on the general topic is here, 
which might get you started on the issues: 
https://www.diva-portal.org/smash/get/diva2:616345/FULLTEXT01.pdf

Conceptually, the problem is that in most situations one needs to have 
some information on the flow "outside of the model domain", because it 
inescapably influences the flow inside (the only exception is 
supercritical outflow). The amount of information that is required 
depends on whether the flow is supercritical or subcritical, and the 
flow is inward or outward from the domain. Note this is a fundamental 
physical issue -- not a matter of numerical implementation.

If you don't provide enough boundary information, then it's hard to say 
what will happen -- maybe the flow will anyway be approximately correct, 
maybe not -- this will vary case-by-case depending on the flow 
situation, and numerical details of the implementation. I expect this is 
happening in your model, at least when using the "Transmissive" boundary 
condition -- unless you have supercritical outflow, there will be 
insufficient information.

Another issue, separate again, is that boundary conditions can also be a 
source of instability, or spurious reflections, depending on the flow 
situation.  And there is yet another issue; the method that ANUGA uses 
to impose boundary conditions will not in general "force" the flow to be 
equal to the boundary value for any quantity -- instead, ANUGA will use 
the boundary values to represent the flow on one "edge" for the flux 
computation.

So there is a lot of potential for trouble with boundary conditions.

That said, if you have a boundary where you expect outflow, and 
furthermore you think the flow is subcritical, then you might consider 
using one of these (see boundaries.py for documentation):
Transmissive_n_momentum_zero_t_momentum_set_stage_boundary
Transmissive_momentum_set_stage_boundary
Time_stage_zero_momentum_boundary
Characteristic_stage_boundary
Flather_external_stage_zero_velocity_boundary

Another good heuristic is to keep your boundary condition "suffiently 
far" from the flow region of interest so that any imperfections do not 
influence the important aspects of the flow. This is not always 
required, but is a good idea if you think you are having trouble at the 
boundary.

Other random techniques -- for some problems you can do things like 
append a "large pond" or similar to trap the outflowing water without it 
causing issues for your domain. Or you can append a very coarse region 
around the boundary, to dissipate important waves. All this is very 
'case-by-case' though.

Good luck,
Gareth.


On 6/5/19 5:48 pm, rupal budhbhatti wrote:
> Hello Sir,
>
> I want the flow out of my domain at the bottom boundary. Correct me If 
> I am wrong but Transmissive boundary is used for that right? Also, 
> Dirichlet boundary condition can be used as drainage using it with 
> parameters as  [0,0,0] that is giving the stage value lower than the 
> lowest elevation.
>
> Sir I am confused between the use of Transmissive and the Dirchlet 
> Boundary condition fused as drainage to get the flow simply out from 
> the domain. I got more confused when I got different results using 
> this two boundary conditions when both are meant for the same purpose. 
> I request you to help me understand where I am wrong.
>
> Please guide me.
>
> Kind Regards,
> Rupal
>
>
>
>
> _______________________________________________
> Anuga-user mailing list
> Anu...@li...
> https://lists.sourceforge.net/lists/listinfo/anuga-user